遊蕩犬族群數量與空間分布的估計，是制定有效族群管理措施的重要基礎。本研究整理一套整合抽樣設計、重捕捉估計與空間預測的方法架構，以評估臺灣遊蕩犬族群數量及分布。首先，本研究回顧 2018 年至 2024 年全國遊蕩犬調查的抽樣設計。調查架構由2018 年的雙重抽樣，演變為 2020 年與 2022 年的二階段抽樣，並於 2024 年進一步採用以村里為單位的一階段分層隨機抽樣。此一調整提升了全國樣本的代表性，並能更細緻地掌握空間異質性。其次，本研究透過模擬，在不同族群規模、偵測機率與調查天數下，比較迴歸法、Petersen–Lincoln 估計量、 Chapman 估計量、Schnabel 及 Schumacher–Eschmeyer 估計量的表現。結果顯示，Schnabel 與 Schumacher–Eschmeyer 估計量在多數情境下具有較低的偏誤與均方根誤差。Chapman 估計量在計算上較為穩定，但隨族群規模增加而呈現低估趨勢；迴歸法則在低偵測機率下較容易出現估計失敗或極端值。將各方法應用於2024年調查資料後，全國遊蕩犬估計數量約介於 117,826 至 346,407 隻之間，顯示估計方法的選擇會顯著影響族群數量推估結果。最後，本研究運用地理加權迴歸預測未抽樣村里的遊蕩犬數量，並探討其與社會經濟及環境因子之間的空間變動關係。結果顯示，人口密度與遊蕩犬數量大致呈負向關聯，而老化指數則呈正向關聯。模型推估全國遊蕩犬總數為 111,521 隻，95%信賴區間為 44,373 至 178,670 隻，並辨識出南臺灣，以及鄉村、沿海、農業與山區等遊蕩犬數量相對較高的地區。整體而言，本研究提出一套整合性方法架構，可用於改善臺灣全國遊蕩犬族群數量估計、空間監測與地方化管理。

Estimation of free-roaming dog (FRD) abundance and spatial distribution is essential for effective group management. This study developed an integrated framework combining sampling design, capture–recapture estimation, and spatial prediction to assess FRD groups in Taiwan.

First, national FRD survey designs implemented from 2018 to 2024 were reviewed. The framework evolved from double sampling in 2018 to two-stage random sampling in 2020 and 2022, followed by one-stage stratified random sampling at the village level in 2024. This refinement improved nationwide representativeness and captured finer-scale spatial heterogeneity.

Second, the regression, Petersen–Lincoln, Chapman, Schnabel, and Schumacher–Eschmeyer estimators were evaluated through simulations under different group sizes, detection probabilities, and survey durations. The Schnabel and Schumacher–Eschmeyer estimators generally produced the lowest bias and root mean square error. The Chapman estimator was computationally stable but tended to underestimate abundance as group size increased, whereas the regression method was vulnerable to failure and extreme estimates under low detectability. Application to the 2024 survey data yielded national estimates ranging from approximately 117,826 to 346,407 dogs, highlighting the influence of estimator selection.

Finally, Geographically Weighted Regression was used to predict FRD abundance in unsampled villages and examine spatially varying relationships with socioeconomic and environmental factors. Population density generally showed a negative association with FRD abundance, whereas the elder index showed a positive association. The model predicted a national total of 111,521 FRDs, with a 95% confidence interval of 44,373–178,670, and identified relatively high-abundance areas in southern Taiwan and in rural, coastal, agricultural, and mountainous regions. Overall, this study provides an integrated framework for improving national FRD estimation, spatial monitoring, and localized management in Taiwan.

[1] Tzu-Yun Yan and Kendy Tzu-yun Teng. Trends in animal shelter management, adoption, and animal death in taiwan from 2012 to 2020, 2023.

[2] Ministry of Agriculture, Executive Yuan, R.O.C. (Taiwan). Farewell to killings: Taiwan implements “no-kill” policy on shelter animals. https://eng.moa.gov.tw/theme_data.php?theme=eng_news&id=481, feb 2017. Accessed: 2026-08-01.

[3] Harish Kumar Tiwari, Abi Tamim Vanak, Mark O’Dea, Jully Gogoi-Tiwari, and Ian Duncan Robertson. A comparative study of enumeration techniques for free-roaming dogs in rural baramati, district pune, india. Frontiers in Veterinary Science, Volume 5 - 2018, 2018.

[4] Vinícius Silva Belo, Guilherme Loureiro Werneck, Eduardo Sergio da Silva, David Soeiro Barbosa, and Claudio Jose Struchiner. Population estimation methods for free-ranging dogs: a systematic review. PLoS One, 10(12):e0144830, 2015.

[5] Wen-Ru Chen. Model-based estimation of the free-roaming dog population size. 2023.

[6] K. Bögel, K. Frucht, G. Drysdale, J. Remfry, World Health Organization, and World Society for the Protection of Animals. Guidelines for Dog Population Management. World Health Organization, 1990.

[7] Helio Shimozako and Euro Couto. Photographic capture-recapture for estimation of stray dog population: A methodological evaluation and an ecological study. 06 2010.

[8] M.H. Kutner. Applied Linear Statistical Models. McGraw-Hill international edition. McGraw-Hill Irwin, 2005.

[9] C. G. Joh. (Carl Georg Johannes) Petersen. On the biology of our flat-fishes and on the decrease of our flat-fish fisheries, with some observations showing how to remedy the latter and promote the flat-fish fisheries in our seas east of the Skaw. Kjøbenhavn Centraltrykkeriet 1894, 1894. https://www.biodiversitylibrary.org/bibliography/52039.

[10] C. Chandra Sekar and W. Edwards Deming. On a method of estimating birth and death rates and the extent of registration. Journal of the American Statistical Association, 44(245):101–115, 1949.

[11] D.G. Chapman. Some Properties of the Hypergeometric Distribution with Applications to Zoological Sample Censuses. University of California publications in statistics. University of California Press, 1951.

[12] G. A. F. Seber. The effects of trap response on tag recapture estimates. Biometrics, 26(1):13–22, 1970.

[13] Norman T. J. Bailey. On estimating the size of mobile populations from recapture data. Biometrika, 38(3/4):293–306, 1951.

[14] Laura Gruber. Capture-recapture methods: A comparison of different estimators and confidence intervals/submitted by laura gruber, bsc. 2023.

[15] Zoe Emily Schnabel. The estimation of the total fish population of a lake. The American Mathematical Monthly, 45(6):348–352, 1938.

[16] CJ Krebs. Estimating abundance in animal and plant populations. Ecological methodology, pages 24–77, 2016.

[17] Alan Beck. The ecology of stray dogs: A study of free-ranging urban animals. Purdue University Press e-books, 01 1973.

[18] Shih-Yuan Fei, Jeng-Tung Chiang, Chang-Young Fei, Chung-Hsi Chou, and Meng-Chih Tung. Estimating stray dog populations with the regression method versus beck’s method: a comparison. Environmental and ecological statistics, 19(4):485–498, 2012.

[19] Schumacher F. X. The estimation of fish populations in lakes and ponds, 1999.

[20] Dan Navarro and Amy Perfors. An introduction to the beta-binomial model. Adelaide: University of Adelaide, 2005.

[21] A.S. Fotheringham, C. Brunsdon, and M. Charlton. Geographically Weighted Regression: The Analysis of Spatially Varying Relationships. Wiley, 2002.

[22] Isabella Gollini, Binbin Lu, Martin Charlton, Christopher Brunsdon, and Paul Harris. Gwmodel: An r package for exploring spatial heterogeneity using geographically weighted models. Journal of Statistical Software, 63(17):1–50, 2015.

[23] Chris Brunsdon, Stewart Fotheringham, and Martin Charlton. Geographically weighted regression-modelling spatial non-stationarity. Journal of the Royal Statistical Society. Series D (The Statistician), 47(3):431–443, 1998.

[24] Yee Leung, Chang-Lin Mei, and Wen-Xiu Zhang. Statistical tests for spatial nonstationarity based on the geographically weighted regression model. Environment and Planning A: Economy and Space, 32(1):9–32, 2000.